Home > APEX > Storage > White Papers > Introduction to APEX File Storage for AWS > Cluster EBS type
APEX File Storage for AWS supports both gp3 EBS volumes for an SSD cluster and st1 EBS volumes for an HDD cluster. These two volume types have different performance characteristics.
The gp3 volume provides a balance of price and performance. It is designed to support a broad range of workloads. By default, a single gp3 volume delivers a consistent baseline throughput performance of 125 MiB/s and IOPS performance of 3000 IOPS. At extra cost, you can provision additional throughput up to 1,000 MiB/s and additional IOPS up to 16,000 IOPS on one gp3 volume independently. This means that you can provision the necessary IOPS or throughput performance without the need for extra capacity. See the AWS documentation General Purpose SSD volumes for more details.
When sizing a cluster, to make the aggregated EBS volume throughput match the instance level EBS bandwidth limit, you have the option to either increase the number of volumes allocated per node with lower throughput performance per volume or decrease the number of volumes allocated per node while increasing the throughput performance per volume. Table 5 shows two examples of an m5dn.12xlarge node configuration by using different gp3 volume configurations.
AWS Instance EBS bandwidth limit | Volume counts per node | Single volume size | Single volume provisioned throughput and IOPS | Aggregated provisioned volume throughput per node | AWS EBS hourly cost per node |
1187.5 MB/s | 10 | 1 TiB | 125 MiB/s | 1250 MiB/s | 1.12 USD |
1187.5 MB/s | 5 | 2 TiB | 250 MiB/s | 1250 MiB/s | 1.16 USD |
Note: the AWS EBS hourly cost is calculated with AWS Pricing Calculator, and assumes the region is us-east-1 for comparison purposes. The cost is the AWS list price and may change.
From the preceding comparison, with the same total raw node capacity, provisioning more volumes with the default EBS volume throughput and IOPS configuration is more cost-effective. In terms of cost, it is advisable to use the default volume throughput and IOPS configuration along with a greater number of volumes when setting up clusters backed by gp3 volumes and having the same capacity and aggregated volume throughput.
The st1 volumes are designed for throughput-intensive workloads such as large, sequential datasets or big data analytics. They consist of magnetic storage that defines performance in terms of throughput rather than IOPS and provides high throughput at a low cost per gigabyte. st1 volumes can also be used for backup and disaster recovery purposes.
St1 volumes use a burst bucket model for performance. Volume size determines both the baseline throughput and burst throughput of your volume. See the AWS documentation Throughput Optimized HDD volumes for more details. When sizing a cluster, to make the aggregated EBS volume throughput match the instance level EBS bandwidth limit, you need to provision sufficient capacity per node.